Gap Navigation Trees for Discovering Unknown Environments

Abstract

We propose a motion planning gap-based algorithms for mobile robots in an unknown environment for exploration purposes. The results are locally optimal and sufficient to navigate and explore the environment. In contrast with the traditional roadmap-based algorithms, our proposed algorithm is designed to use minimal sensory data instead of costly ones. Therefore, we adopt a dynamic data structure called Gap Navigation Trees (GNT), which keeps track of the depth discontinuities (gaps) of the local environment. It is incrementally constructed as the robot which navigates the environment. Upon exploring the whole environment, the resulting final data structure exemplifies the roadmap required for further processing. To avoid infinite cycles, we propose to use landmarks. Similar to traditional roadmap techniques, the resulting algorithm can serve key applications such as exploration and target finding. The simulation results endorse this conclusion. However, our solution is cost effective, when compared to traditional roadmap systems, which makes it more attractive to use in some applications such as search and rescue in hazardous environments.